Techniques for use of carbon nanotubes as an anti-tampering feature and for use of randomly metallic or semiconducting carbon nanotubes in the generation of a physically unclonable cryptographic key generation are provided. In one aspect, a cryptographic key having an anti-tampering feature is provided which includes: an array of memory bits oriented along at least one bit line and at least one word line, wherein each of the memory bits comprises a memory cell, wherein the cryptographic key is stored in the memory cell, and wherein the memory cell is connected to the at least one bit line; and a metallic carbon nanotube interconnect which connects the memory cell to the at least one word line. A cryptographic key and method for processing the cryptographic key are also provided.
|
1. A method for processing a cryptographic key, the method comprising the steps of:
forming the cryptographic key which comprises an array of memory bits oriented along at least one bit line and at least one word line, wherein each of the memory bits comprises a selecting transistor having a source and a drain interconnected by a channel and a gate over the channel, wherein the gate is connected to the at least one word line, and wherein for at least one of the memory bits in the array the source is connected to the at least one bit line via either a metallic carbon nanotube or a semiconducting carbon nanotube, wherein the forming step further comprises: randomly depositing either the metallic carbon nanotube or the semiconducting carbon nanotube to form an interconnect between the source and the at least one bit line, whereby a randomness of whether the interconnect of a given one of the memory bits is formed from the metallic carbon nanotube or the semiconducting carbon nanotube is the cryptographic key;
reading an output state of each of the memory bits; and
storing the output state of each of the memory bits in a server.
2. The method of
blowing the e-fuse once the output state of each of the memory bits has been read.
3. The method of
forming trenches in the dielectric material adjacent to the at least one bit line; and
placing the carbon nanotube within one of the trenches.
4. The method of
5. The method of
forming the trenches in the SiO2 layer such that the HfO2 layer remains present within the trenches; and
selectively forming a monolayer of NMPI within the trench on the HfO2 layer.
6. The method of
preparing an aqueous solution of carbon nanotubes comprising both metallic carbon nanotubes and semiconducting carbon nanotubes, wherein the carbon nanotubes are wrapped in a surfactant; and
contacting the aqueous solution of the carbon nanotubes with the monolayer of NMPI such that by way of attractive forces between the NMPI monolayer which is positively charged and the surfactant which is negatively charged the carbon nanotubes self-assemble within the trenches, and wherein the carbon nanotubes that self-assemble within the trenches are randomly either the metallic carbon nanotubes or the semiconducting carbon nanotubes.
7. The method of
8. The method of
|
|||||||||||||||||||||||||||||
This application is a continuation of U.S. application Ser. No. 14/586,103 filed on Dec. 30, 2014, the disclosure of which is incorporated by reference herein.
The present invention relates to cryptographic key generation and protection and more particularly, to techniques for use of carbon nanotubes as an anti-tampering feature and for use of randomly metallic or semiconducting carbon nanotubes in the generation of a physically unclonable cryptographic key generation.
A unique cryptographic key stored securely in the server is used to verify the authenticity of a chip (i.e., to determine whether a chip is real or fake). Specifically, during chip authentication the server uses the serial ID number of the chip to pull up the specific, unique key associated with that chip. The server then sends a randomly generated challenge question (e.g., some code) to the chip which the chip then encrypts and sends back to the server. The chip typically uses a universal encryption algorithm coupled with its unique key. The server then verifies the ID of the chip by decrypting the output from the chip using the key the server has associated with that chip ID. The server compares the decrypted challenge question with the original challenge question sent to the chip. Only if the two match can the chip be authenticated.
Protecting the key is one area of vulnerability in typical systems. For instance, as encryption/decryption is usually done by the application-specific integrated circuit (ASIC), cryptographic keys (unencrypted) often appear in data random access memory (RAM), registers and cache memory. All of these storage elements have transistors switching whenever a value of data is changed. Switching of transistors causes information leakage through various channels including power supply line and electromagnetic emission. Current consumption analysis can be accessed through direct probing of the circuit power line, which requires reverse-engineering of the chip.
It is also possible to direct access word line (WL)/bit line (BL) of cryptographic key arrays in the chip to steal the key by reverse-engineering. Once the key is stolen, hackers can duplicate unlimited number of chips and the chip authentication is compromised.
Furthermore, cryptographic keys based on charge-storing memories are usually vulnerable to Differential Power Analysis (DPA). For example, the current consumption difference before and after baking the chip at a high temperature (which shifts charge stored) can be easily used to estimate “1” and “0” in the key. Therefore, it is highly desirable to generate the random key using non charge-storing based technology.
Accordingly, techniques for cryptographic key generation and protection which are more robust against reverse-engineering and other procedures used to impermissibly obtain the key would be desirable.
The present invention provides techniques for use of carbon nanotubes as an anti-tampering feature and for use of randomly metallic or semiconducting carbon nanotubes in the generation of a physically unclonable cryptographic key generation. In one aspect of the invention, a cryptographic key having an anti-tampering feature is provided. The cryptographic key having the anti-tampering feature includes: an array of memory bits oriented along at least one bit line and at least one word line, wherein each of the memory bits comprises a memory cell, wherein the cryptographic key is stored in the memory cell, and wherein the memory cell is connected to the at least one bit line; and a metallic carbon nanotube interconnect which connects the memory cell to the at least one word line, wherein the metallic carbon nanotube interconnect is the anti-tampering feature which is configured to be removed during a reverse-engineering process and with it any connection between the memory cell and the at least one word line.
In another aspect of the invention, a cryptographic key is provided. The cryptographic key includes: an array of memory bits oriented along at least one bit line and at least one word line, wherein each of the memory bits has a selecting transistor having a source and a drain interconnected by a channel and a gate over the channel, wherein the gate is connected to the at least one word line, and wherein for at least one of the memory bits in the array the source is connected to the at least one bit line via either a metallic carbon nanotube or a semiconducting carbon nanotube.
In yet another aspect of the invention, a method for processing a cryptographic key is provided. The method includes the steps of: forming the cryptographic key which comprises an array of memory bits oriented along at least one bit line and at least one word line, wherein each of the memory bits comprises a selecting transistor having a source and a drain interconnected by a channel and a gate over the channel, wherein the gate is connected to the at least one word line, and wherein for at least one of the memory bits in the array the source is connected to the at least one bit line via either a metallic carbon nanotube or a semiconducting carbon nanotube; reading an output state of each of the memory bits; and storing the output state of each of the memory bits in a server.
A more complete understanding of the present invention, as well as further features and advantages of the present invention, will be obtained by reference to the following detailed description and drawings.
Provided herein are techniques for cryptographic key generation and protection for chip authentication based on carbon nanotube technology. The cryptographic keys provided herein are more robust against reverse-engineering and other tactics used to impermissibly obtain the key.
Carbon Nanotube-Based Cryptographic Key Anti-Tampering Devices—
In a first exemplary embodiment of the present techniques, metallic carbon nanotube interconnects are employed as part of a cryptographic key anti-tampering device. In the example now presented, the metallic carbon nanotube forms part of a word line (WL) connection to a memory cell—wherein the cryptographic key is stored in the memory cell. The metallic carbon nanotube prevents any reverse-engineering to directly access any metal levels below the metallic carbon nanotube. Specifically, as will be described in detail below, with this exemplary configuration the word line signal will pass through higher metal levels and the short metallic carbon nanotube segment, and return to the memory cell. Attempts to reverse engineer by delayering the device will destroy the metallic carbon nanotube interconnect, and thereby the stored key will lose its connection and function. Namely, the way often employed to obtain a key via reverse engineering is to electrically measure each bit (1 or 0). In order to do so, a hacker has to remove the upper dielectric and metal levels (delayering) until he/she can electrically probe the lead. Such attempts would be thwarted by the present anti-tampering features.
An exemplary process for forming a cryptographic key having a metallic carbon nanotube-based anti-tampering feature according to this exemplary embodiment is described by way of reference to
In general, the carbon nanotube-based anti-tampering feature will be used to form an interconnect between the word line (or higher metal level) and the memory cell. What is important is that this carbon nanotube anti-tampering feature is destroyed (during the above-described delayering process) before the word line can be accessed.
The example illustrated in the figures uses a standard memory transistor such as a floating-gate transistor which includes a source (S) and a drain (D) interconnected by a channel, and a gate (G) over the channel which serves to regulate electron flow through the channel. Spacers are often used on opposite sides of the gate (G) to offset the gate (G) from the source (S) and the drain (D). See
In the floating-gate transistor example, the transistor may be fabricated using standard processing techniques including, but not limited to, defining an active area of the memory cell in a wafer, gate (G) material deposition and patterning, spacer formation, source (S) and drain (D) implantation, etc. The process for fabricating a transistor on a wafer would be apparent to one skilled in the art, and thus is not described further herein. It is noted that for ease and clarity of depiction, the wafer (substrate upon which the memory transistor is constructed) is not shown. However, any suitable semiconductor wafer configuration may be employed including, but not limited to, a bulk semiconductor wafer, a semiconductor-on-insulator (SOI) wafer, etc.
In this exemplary embodiment, conventional key storage technology is employed in the sense that the key is stored in the memory cell (e.g., in this example—in the memory transistor). Advantageously, however, the present techniques introduce a unique metallic carbon nanotube-based anti-reverse engineering feature. As will be described in detail below, when this metallic carbon nanotube-based anti-reverse engineering feature is destroyed during a delayering process aimed to extract the key, the stored key will lose its connection and function. By contrast, in another exemplary embodiment provided below, carbon nanotube-based technology is used to physically create the cryptographic key itself, i.e., randomly deposited metallic or semiconducting carbon nanotubes become the key.
As provided above, the memory cell is accessed via a word line (WL) and a bit line (BL), wherein an intersection of the word line and the bit line is the address of the memory cell. In the instant example of a floating-gate transistor as the memory cell, the word line contacts the gate (G) of the transistor. Thus, as will be described in detail below, this provides an ideal location for implementation of the metallic carbon nanotube anti-tampering feature since delayering reverse engineering efforts (which deconstruct the device layer-by-layer typically from the top down) will remove the metallic carbon nanotube and thereby the stored key will lose its gate connection and function. As would be apparent to one skilled in the art, other memory cell configurations which do not have a gate-to-word line connection still provide the opportunity to insert the present anti-tampering feature in a metal layer above the word line (such that upon delayering—by the time the word line is accessed the connection to the word line is removed). By way of example only, in the case of MRAM, each magnetic memory cell is accessed via a word line and a bit line running above and below (or alternatively below and above) the memory cell. See, for example, U.S. Pat. No. 7,433,225 issued to Daniel C. Worledge, entitled “Scalable magnetic random access memory device,” the contents of which are incorporated by reference as if fully set forth herein. Thus, the implementation of MRAM as opposed to a memory transistor would provide the same word line connection to the memory cell (or other higher metal layer) into which the present carbon nanotube interconnect-based anti-tampering feature could be placed. Thus, while the description that follows highlights an implementation involving a memory transistor one skilled in the art could apply this teaching to place the present anti-tampering feature in the metal layers accessing any type of non-volatile memory cell.
The process for forming the word and bit lines to include the metallic carbon nanotube anti-tampering feature will now be described. Specifically, as shown in
In this example, the source (S), drain (D), and gate (G) contacts consist of metal (or other electrically conductive material)-filled vias that connect the memory transistor to a first metal layer (M1). As is known in the art, the process for forming source (S), drain (D), and gate (G) contacts involves depositing a dielectric material (DE1) (such as silicon dioxide (SiO2)) over the memory transistor, patterning vias in the dielectric material (DE1), and then filling the vias with a conductive material such as a metal or metals. For ease and clarity of depiction, a simple dotted line is being used to schematically represent the placement of the dielectric material (DE1) with the understanding that the configuration of such a standard interlayer dielectric would be apparent to one skilled in the art.
Next, a M1 metal layer is formed contacting the source (S) and gate (G) contacts. See
M1-M2 interconnects are then formed which will serve to interconnect the M1 metal layer with an M2 metal layer (see below). See
Finally, an M2 metal layer is formed. See
As highlighted above, with a conventional memory bit the word line (WL) is connected directly to the gate (G). By contrast, in accordance with the present techniques, the metallic carbon nanotube-based anti-tampering feature is inserted between the word line (WL) and the gate (G). Carbon nanotubes are small and thus cannot be easily detected without advanced imaging techniques such as a scanning electron microscope (SEM). Thus, if one were to attempt to reverse-engineer the present device by polishing down the bit layer-by-layer (i.e., a delayering process), by the time the word line (WL) is reached there will be no visible connection between the word line (WL) and the gate since the metallic carbon nanotube interconnect will have to be removed before one even reaches the word line (WL).
It is notable that, as highlighted above, the carbon nanotube used in this example is a metallic carbon nanotube. Carbon nanotubes, as produced, will naturally contain both semiconducting and metallic carbon nanotubes. For use as an interconnect, a metallic carbon nanotube is needed. Techniques are known in the art for separating metallic from semiconducting carbon nanotubes which may be implemented in accordance with the present techniques. See, for example, Tulevski et al., “High Purity Isolation and Quantification of Semiconducting Carbon Nanotubes via Column Chromatography,” ASC Nano, vol. 7, no. 4, pgs. 2971-2976 (March 2013) (hereinafter “Tulevski”), the contents of which are incorporated by reference as if fully set forth herein. Using the techniques described in Tulevski, sorted samples of metallic carbon nanotubes can be obtained (see, for example, FIG. 1 of Tulevski).
The structure shown in
The general implementation of the metallic carbon nanotube interconnect anti-tampering feature was described above. Now provided is an exemplary technique that might be implemented in accordance with the present techniques to selectively place the present metallic carbon nanotube interconnect within the M2 metal layer. In general, any technique used for the selective placement of a metallic carbon nanotube(s), in a specified position for use as an interconnect may be employed in accordance with the present techniques. For instance, metallic carbon nanotubes can be simply cast from an aqueous solution (for example using a spin-coating process) to place the present metallic carbon nanotube interconnect. A mask may be employed to ensure proper placement. By way of example only, according to one non-limiting exemplary embodiment, chemical self-assembly techniques are used as described, for example, in Park et al. “High-density integration of carbon nanotubes via chemical self-assembly,” Nature Nanotechnology, 7, 787-791 (October 2012) (hereinafter “Park”), the contents of which are incorporated by reference as if fully set forth herein. As described in Park, selective placement can be achieved based on ion exchange between a functional surface monolayer and surfactant-wrapped carbon nanotubes which are processed in an aqueous solution. Specifically, a surface monolayer is formed from 4-(N-hydroxycarboxamido)-1-methylpyridinium iodide (NMPI) which contains hydroxamic acid end groups that are known to self-assemble on metal oxide surfaces (such as HfO2) but not on SiO2. The functionalized surface can then be contacted with an aqueous solution of surfactant-wrapped carbon nanotubes whereby the anion of the NMPI is exchanged with the anionic surfactant wrapped around the carbon nanotubes leading to a strong coulombic attraction between the negatively charged surfactant and the positively charged monolayer. As a result, self-assembly of the carbon nanotubes can be achieved on the HfO2 surfaces selective to the SiO2 surfaces. An exemplary application of this process described in Park to the present techniques is now described by way of reference to
Since the placement of the metallic carbon nanotube interconnect occurs (according to this example) during the formation of the M2 metal layer (see above), the description starts at
The metallic carbon nanotube interconnect is then placed followed by formation of the M2 metal lines. Specifically, as mentioned above, the metallic carbon nanotube interconnect is placed (selectively) within a trench formed in the dielectric. Specifically, standard lithography and etching can be used to form a trench 902 in the SiO2 layer 802. See
An aqueous solution of surfactant-wrapped metallic carbon nanotubes is prepared. According to an exemplary embodiment, the carbon nanotubes are wrapped by a surface-active dispersing agent, such as sodium dodecylsulfate (SDS). Wrapping the carbon nanotubes in the surfactant allows the carbon nanotubes to be dispersed in water forming the aqueous solution. Column chromatography is then used to isolate the metallic carbon nanotubes according to the process described in Tulevski. Excess SDS can be removed by dialysis to obtain a high carbon nanotube deposition yield. The aqueous solution of the surfactant-wrapped metallic carbon nanotubes is then contacted with the NMPI surface. See
The M2 metal lines can then be formed in contact with the metallic carbon nanotube interconnect and with the underlying M1-M2 interconnects. Specifically, standard lithography and etching can be used to form a pattern 1202 for the M2 metal layer in the bilayer dielectric. See
The pattern 1202 is then filled with a conductive material such as a metal (e.g., copper)—forming the M2 metal lines. See
The above-described configuration wherein the metallic carbon nanotube interconnecting the word line (WL) and the gate is placed in the M3 metal layer is ideal in terms of reverse-engineering proofing the circuit. Namely, when delayering the circuit from the top down (it is common to remove layers one by one starting from the top) by the time the word line (WL) is reached there will be no visible connection to the gate. It is notable that other configurations are however possible. For instance, in accordance with the present techniques, a metallic carbon nanotube interconnect may instead (or in addition to) be placed in one or more higher metal levels (so long as the carbon nanotube anti-tampering feature is destroyed (during reverse-engineering efforts) before the word line is revealed—thereby removing any connection between the memory cell and the word line. Redundancy can increase the degree of protection. The same process described above for selectively placing metallic carbon nanotube(s) can be applied to any layer in the circuit. For example, a (HfO2/SiO2) bilayer dielectric can be employed for any of the dielectric layers (DE1-4) thus enabling selective carbon nanotube placement via the self-assembly process in Park. See above.
Physically Unclonable Cryptographic Keys Based on Self-Assembled Carbon Nanotubes—
In the embodiments described thus far, only the use of metallic carbon nanotubes has been desirable as their conductivity is necessary to form an interconnect. As provided above, however, as-synthesized carbon nanotubes contain both metallic and semiconducting carbon nanotubes. Specifically, due to different chiralities about ⅓ of carbon nanotubes exhibit metallic behaviors, and about ⅔ show semiconducting behaviors. Apart from their very different conductance, these two types of carbon nanotubes are basically indistinguishable from one another. Advantageously, provided herein are techniques that leverage this randomly varying conductance property of carbon nanotubes to create physically unclonable cryptographic keys.
Specifically, the key is generated during fabrication of the circuit by depositing randomly either metallic or semiconducting carbon nanotubes. Thus, the randomness of the deposited carbon nanotubes (as either semiconducting or metallic) is the “key” itself. As will be described in detail below, once the key is generated during fabrication of the circuit it can be read from the chip and stored by the server. By comparison, in the embodiment described in conjunction with the description of
The same chemical self-assembly process as described above is used to selectively place the carbon nanotubes. Here however the use of both metallic and semiconducting carbon nanotubes adds a level of randomness to the process based on the inherent physical disorder of the chemical self-assembly process. Such randomness is desirable for cryptographic key generation since it makes reverse-engineering the key impossible, even if one were to know the exact process by which it was made. Another level of randomness is provided based on the width of the trenches into which the carbon nanotubes are deposited. By varying the trench width, the possibility of a carbon nanotube being deposited into the trench can be controlled. For instance, a larger trench width increases the chance that a carbon nanotube is deposited into the trench. Conversely, a smaller trench width can reduce the chance of carbon nanotube deposition. Therefore, each bit in a carbon nanotube cryptographic key can be metallic, semiconducting, or no-tube. This process provides a low-cost and hard-to-forge approach for implementing physical security primitives.
As compared to the above approach wherein carbon nanotube-based anti-tampering techniques are implemented in a conventional cryptographic key configuration, inherent physical disorder is now being leveraged as a hardware roots-of-trust cryptographic key. Hardware roots-of-trust cryptographic keys are described, for example, in Tehranipoor et al., “Introduction to Hardware Security and Trust,” Springer: New York, N.Y., 2012 (hereinafter “Tehranipoor”), the contents of which are incorporated by reference as if fully set forth herein. Physical one-way functions are described, for example, in Pappu et al., “Physical One-Way Functions,” Science, vol. 297, pgs. 2026-2030 (September 2002) (hereinafter “Pappu”), the contents of which are incorporated by reference as if fully set forth herein. As described in Pappu, a unique speckle pattern can be generated when shining a laser through an optical epoxy token with inhomogeneous scatters. The format of the physical unclonable function (PUF) was later expanded from non-electronic devices to integrated circuit-based electronic devices. See, for example, Gassend et al., “Silicon Physical Random Functions,” Proceedings of the 9th ACM conference on Computer and communications security CCS '02, pgs. 148-160 (November 2002) (hereinafter “Gassend”), and Maes et al., “Physically Unclonable Functions: A Study on the State of the Art and Future Research Directions,” Towards Hardware-Intrinsic Security, Information Security and Cryptography 2010, pp. 3-37 (October 2010), the contents of each of which are incorporated by reference as if fully set forth herein. All silicon PUFs exploit local mismatches between different circuit components. As a major component, metal-oxide-semiconductor field effect transistors (MOSFET) are subjected to the mismatches from random dopant fluctuation in the channel, line edge roughness and polysilicon/high-K granularity. The mismatches arising from these stochastic atomic variations are likely to follow a Gaussian distribution, in which a large portion of the mismatches are very close to zero mean. PUFs based on these small mismatches can be easily disturbed by noises and environmental factors which can lead to large bit error rates in the output. Thus pre-processing, to rule out devices with small mismatches or post-processing to reduce the bit error rate is required in silicon PUFs. For example, ring oscillator or arbiter PUFs, are based on mismatches of two or multiple delay chains, which are subject to the averaging effect of mismatches in multiple stages. See, for example, S. S. Mansouri et al., “Ring Oscillator Physical Unclonable Function with Multi Level Supply Voltages,” 2012 IEEE 30th International Conference on Computer Design (ICCD), pgs. 520-521 (September/October-2012) (ring oscillator), the contents of which are incorporated by reference as if fully set forth herein, and Gassend (arbiter PUFs).
However, these traditional complementary metal-oxide-semiconductor (CMOS) based security primitives not only require large amounts of dedicated logic circuits, area, and power, but they also require high accuracy measurements utilizing high frequency clocks and high resolution timing. See, for example, Beckmann et al., “Hardware-Based Public-Key Cryptography with Public Physically Unclonable Functions,” Lecture Notes in Computer Science, vol. 5806, pp. 206-220 (2009), the contents of which are incorporated by reference as if fully set forth herein. Nanotechnology enables conceptually new security primitives which are potentially more robust and tamper-resistant than CMOS based security primitives. See, for example, Rose et al., “Hardware Security Strategies Exploiting Nanoelectronic Circuits,” 2013 18th Asia and South Pacific Design Automation Conference (ASP-DAC), pp. 368-372 (January 2013), and Rose et al., “Nanoelectronics and Hardware Security,” Network Science and Cybersecurity, Advances in Information Security, vol. 55, pp. 105-123 (2014), the contents of each of which are incorporated by reference as if fully set forth herein. Most of the existing functional nanomaterials exploit their unique optical properties for anti-counterfeiting applications, either by information encryption or naturally occurring randomness. See, for example, B. Yoon et al., “Recent functional material based approaches to prevent and detect counterfeiting,” J. Mater. Chem. C, 1, 2388-2403 (January 2013) (information encryption), Demirok et al., “Orthogonal Multi-Readout Identification of Alloy Nanowire Barcodes,” J. AM. CHEM. SOC. 131, 22-23 (December 2008) (naturally occurring randomness), and Kim et al., “Anti-counterfeit nanoscale fingerprints based on randomly distributed nanowires,” Nanotechnology 25, 155303 (March 2014) (naturally occurring randomness), the contents of each of which are incorporated by reference as if fully set forth herein. For the ease of integration into a system, particularly for the on-chip security, electronic devices with easy access are more favorable. However, currently only few examples exist, such as the memristor-based PUF which possesses merits of high density and non-volatility but requires SET-RESET steps. See, for example, Rajendran et al., “Nano-PPUF: A Memristor-Based Security Primitive,” 2013 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), pp. 84-87 (August 2012), the contents of which are incorporated by reference as if fully set forth herein.
Provided herein is a non-volatile, low-cost PUF (also referred to herein as a “physically unclonable function”) which is based on self-assembled carbon nanotubes. In general, the present physically unclonable cryptographic key includes an array of memory cells. Each of the memory cells represents a single bit. Thus, the terms “memory cell” and “bit” are used interchangeably herein. In the example described immediately below, each bit includes a selecting field effect transistor (FET) (or simply “selecting transistor”). For smaller arrays a selecting transistor may not be required, and such an example is provided below. In most cases however, with larger arrays selecting transistors are needed to access specific bits. Generally, the implementation of selecting transistors in a memory cell array should be apparent to one skilled in the art.
An exemplary methodology for fabricating the bits in the array is now provided by way of reference to
The process begins by forming the selecting FET (one for each of the memory cells). See
Each bit is accessed via a word line (WL) and a bit line (BL), wherein an intersection of the word line and the bit line is the address of the bit. In most configurations, the word line contacts the gate (G) of the memory cell and the bit line contacts the source (S)/drain (D). Thus, it is through the bit lines which information is written to/read from the bits. As will be described in detail below, a carbon nanotube(s) is inserted in the bit line connection to the source/drain via the above-described self-assembly process which is randomly metallic or semiconducting. Additionally, by varying the width of the trenches into which the carbon nanotubes are deposited, some bits may receive no carbon nanotube at all, thus adding to the randomness of the key generation process.
The process for forming the word and bit lines and the (metallic or semiconducting) carbon nanotube key will now be described. Specifically, as shown in
In this example, the source (S), drain (D), and gate (G) contacts consist of metal (or other electrically conductive material)-filled vias that connect the selecting transistor to a first metal layer (M1). As is known in the art, the process for forming source (S), drain (D), and gate (G) contacts involves depositing a dielectric material (DE1) (such as SiO2) over the selecting transistor, patterning vias in the dielectric material (DE1), and then filling the vias with a conductive material such as a metal or metals. For ease and clarity of depiction, a simple dotted line is being used to schematically represent the placement of the dielectric material (DE1) with the understanding that the configuration of such a standard interlayer dielectric would be apparent to one skilled in the art.
Next, an M1 metal layer is formed contacting the source (S) and gate (G) contacts. See
M1-M2 interconnects are then formed which will serve to interconnect the M1 metal layer with an M2 metal layer (see below). See
Finally, an M2 metal layer is formed. See
In order to place the carbon nanotube(s) in the M2 metal layer, a trench is formed in the DE4 dielectric material. Based on the self-assembly process described for example in conjunction with the description of
An aqueous solution of SDS-wrapped carbon nanotubes is prepared. Excess SDS can be removed by dialysis to obtain a high carbon nanotube deposition yield. Based on the different chiralities of the carbon nanotubes as synthesized, the solution will contain both metallic and semiconducting carbon nanotubes. In this case, it is desirable to have both metallic and semiconducting carbon nanotubes since the randomness of placing either a metallic or semiconducting carbon nanotube in a given bit based on the self-assembly process contributes to the ‘unclonability’ of the key. Namely, by contacting the carbon nanotube solution with the NMPI coated surface the carbon nanotubes will self-assemble in the trenches (throughout the array) due to the strong Coulombic attraction between the positively charged (NMPI) monolayer and the negatively charged SDS (see above). Thus, even if one were to know the exact process by which the bits were created, the randomness of the (metallic and/or semiconducting) carbon nanotube placement makes reproducing the key impossible (i.e., unclonable). As will be described in detail below, in accordance with the present techniques, once created the key can be read out once and saved to the server.
The structure shown in
Once the key has been created, it can be read from the chip and stored in the server.
For enhanced security, the readout pads can optionally be connected to the bits via electronic fuses (e-fuses). For a description of e-fuses, see for example, U.S. Pat. No. 8,189,419 issued to Chen et al., entitled “Apparatus for nonvolatile multi-programmable electronic fuse system,” the contents of which are incorporated by reference as if fully set forth herein. Once the key readout has occurred in step 1904, the e-fuses can be blown which means that the key (the carbon nanotube array information) cannot be directly accessed anymore—the output of the key array can only enter the on-die encryption circuit. For a description of on-die encryption/decryption circuits see, for example, Krishnamurthy et al., “High-performance energy-efficient encryption in the sub-45 nm CMOS era,” 2011 48th ACM/EDAC/IEEE Design Automation Conference (DAC) (June 2011), the contents of which are incorporated by reference as if fully set forth herein.
Finally, in step 1906 the key readout from the chip is stored to the server. As shown in
As provided above, the dimensions of the trenches can be varied to further increase the randomness of the key. By way of example only, a trench width can be chosen (i.e., the same width will be used for all the trenches on a given chip) that provides about a 50% chance that a carbon nanotube will be deposited inside the trench. Since as will be described in detail this placement process is based on both attractive and repulsive forces acting on the carbon nanotube, the placement process is also random (one cannot control which trenches will eventually have a carbon nanotube) and thus tailoring the trench width in this manner can create additional randomness. For instance, as will be described in detail below, reducing the trench width increases the interaction between the SiO2 sidewalls of the trench and the surfactant-wrapped carbon nanotubes, both of which are negatively charged—generating repulsive forces. Reducing the trench width can thus lead to the absence of carbon nanotubes in one or more of the bits. See, for example,
Thus, the width of patterned HfO2 trenches can be varied, if so desired, to achieve random placement of the carbon nanotubes. This concept can be illustrated by the following non-limiting example—by reducing the trench width from about 300 nanometers (nm) to about 70 nm, the repulsive force between negatively charged SiO2 sidewall and negatively charged SDS wrapping around the carbon nanotube (SDS-CNT) becomes more prominent compared to the attractive force between the NMPI monolayer and the SDS-CNT. By carefully designing the trench dimension, this competition between the attractive force and the repulsive force can lead to highly random CNT placement inside the trench, see
As shown in
There are several potential approaches for harvesting the inherent randomness of self-assembled carbon nanotube bits. By setting different threshold current, a unique distribution can be obtained for each current level. A ring oscillator or an arbiter (see, for example, Suh et al., “Physical unclonable functions for device authentication and secret key generation,” Proceedings of the 44th annual Design Automation Conference DAC '07, pgs. 9-14 (June 2007), the contents of which are incorporated by reference as if fully set forth herein) can also be built using carbon nanotubes instead of silicon to make lower power devices. These approaches however require sophisticated engineering control of carbon nanotubes and the fabrication process. A simpler and reliable method involves determining the connection type of nanotube devices. Given the stability of CNTs over a long period of time (see, for example, Shahrjerdi et al., “High-Performance Air-Stable n-Type Carbon Nanotube Transistors with Erbium Contacts,” ACS Nano, 2013, 7(9), pp. 8303-8308 (September 2013), the contents of which are incorporated by reference as if fully set forth herein) and a large noise margin between connected and disconnected devices, this approach provides exceptional repeatability.
To understand the trench-width dependent carbon nanotube placement behavior, numerical calculations (see below) may be made based on a multiphysics model including electrostatics and ionic transport. A description of the multiphysics model is now provided.
Simulation—a simulation of the interaction between SDS-wrapped carbon nanotubes and the charged surface was carried out using COMSOL Multiphysics® Modeling Software based on finite element methods. The structure is shown in
The possible combination number—in an array with a total device number of n, the possible combination number C2(n,m) of connected devices (m) and open devices (n−m) is given by the combination:
In order to evaluate the dependence of C2(n,m) on m and n, Stirling's approximation given by:
ln(n!)≈n ln(n)−n (2)
is applied, and it is assumed that m=αnα∈(0,1]. The simplified form of the combination equation can be represented by:
ln(C2(n,m))≈−n[(1−α)ln(1−α)+α ln(α)]≡−nγ(α). (3)
By taking a derivative of function γ(α), the minimum value of γ(α)=−0.6931 can be calculated when α=0.5. And the function ln C2(n,m) is symmetric about axis of α=0.5, as shown in
Based on this multiphysics model, the following numerical calculations are used to analyze the trench-width dependent carbon nanotube placement behavior:
(4) Poisson equation describes the electric potential based on charge distribution:
−∈0∇·[∈(x)∇Ψ(x)]=ρ(x)=e0NAΣi=1N
wherein, Ψ(x) is the electric potential, ρ(x) is the net charge density, ∈0 is the vacuum permittivity and ∈(x) is the dielectric constant of the solution. ci(x) and Zi denote the molar concentration and valence of the ith ionic species respectively.
(5) Nernst-Planck equation describes the ion motion in the solution:
Ji(x)=ci(x)ν(x)−Di∇ci(x)−μici(x)∇Ψ(x). (5)
The diffusion coefficient Di is related to the mobility μi via Einstein relation, μi=Zie0Di/KBT. ν(x) is the velocity field of the fluid. The equation is solved under steady state, where ∇·Ji(x)=0.
Although the interaction between a carbon nanotube and the substrate surface can be a complicated three-dimensional problem, it can be simplified by using a two-dimensional simulation by assuming that the carbon nanotube is always parallel to the elongated direction of the trench. This is a reasonable assumption since the minimum energy state occurs when the carbon nanotube locates near the center of the trench, as will be shown below. Therefore, the carbon nanotube will be forced to align along the elongated direction of the trench as long as the trench has a high length-to-width aspect ratio. Two geometries with trench width/barrier width of 30 nm/30 nm and 80 nm/80 nm were simulated and the resulting electric potential maps are plotted in
To show that the carbon nanotube aligns along the elongated direction of the placement trench, the case when moving the carbon nanotube in parallel with the surface at a fixed separation was also simulated. An electric potential map of the 30 nm/30 nm geometry with CNT position at (x=−15 nm, y=25 nm) is shown in
As provided above, for larger arrays it is preferable that each of the bits in the array has a selecting FET. The FETs enable accurate selection of any given bit in the array. If the array is small enough, however, it is possible to construct bits without a selecting FET. An example of a simplified small array of bits that do not include selecting FETs is shown in
According to an exemplary embodiment, this two-dimensional array of carbon nanotube devices is fabricated by the following process. Carbon nanotubes are first deposited using the above-described chemical self-assembly method. In this example, trenches with different dimensions are patterned on an HfO2 blanket film, followed by 7 nm thick SiO2 evaporation and a lift-off process. Electron-beam lithography is used to define all the patterns mentioned here, either in poly(methyl methacrylate) (PMMA) or hydrogen silsesquioxane (HSQ). Right before self-assembly of the NMPI monolayer, the substrate is cleaned in oxygen plasma (0.3 torr, 5 minutes). The patterned substrate is then immersed in a 3.5 mM solution of NMPI (3:1 ethanol/water) for 1 hour, followed by rinsing with ethanol to complete the NMPI monolayer coating. Several drops of nanotube solution are put down on top of the substrate and covered with a thin glass slide. After nanotube deposition for 1 hour, the substrate is rinsed with deionized (DI) water for 30 seconds and sonicated in DI water for 1 minute. The carbon nanotubes can be patterned by reactive ion etch (RIE) in oxygen using double layers of PMMA and HSQ as a etch mask. The HSQ layer is removed by lifting off underlying PMMA layer in hot acetone. The substrate is then thermal annealed at 400 degrees Celsius (° C.) in vacuum of 1×10−7 torr to drive off the monolayer and SDS to recover the electrical properties of CNTs. The first metal layer (0.5 nm titanium (Ti)/20 nm palladium (Pd)/10 nm gold (Au)) is fabricated using a lift-off process as source and drain electrodes. 2% HSQ is patterned as an insulating layer between two metal layers. The second metal layer (0.5 nm Ti/20 nm Pd/30 nm Au) is then patterned as wires and pads. The completed device can be annealed at 150° C. in hexamethyldisilazane (HMDS) ambient for 40 minutes to improve device performance. The highly doped silicon substrate serves as a backgate.
For an array of a given size, the number of connected units (“1”) and disconnected units (“0”) should be equal in order to achieve the maximum combination randomness (see above). Therefore, according to an exemplary embodiment, the trench width of 80 nm is chosen for its ability to realize a connection yield close to 50%. By way of example only,
Synthesized carbon nanotubes naturally contain both semiconducting and metallic nanotubes, which is detrimental to electronic applications but beneficial to cryptographic keys application. By setting an additional threshold current of 10 nA at gate voltage of 0 V, one is able to discriminate between switching and non-switching devices among those connected devices. 2-state random bits can thus be upgraded to 3-state random bits (
One key feature of an ideal cryptographic key technology is the capability to prevent any unauthorized duplication. The carbon nanotube key possesses physical unclonability originated from the randomness of nanoscale diffusion and drift process. Even knowing the same fabrication process, it is impossible to clone the same carbon nanotube key. Optical methods do not have enough resolution to detect the presence of a single nanotube in the trench. Electron microscope imaging requires chip reverse-engineering process, which can easily destroy the nanotube. Attacks by electrical measurement are possible only if knowing the lead layout. In addition, utilizing the switching behavior of semiconducting nanotubes by incorporating local gates as additional inputs makes the already difficult process even more daunting. As shown in
Although illustrative embodiments of the present invention have been described herein, it is to be understood that the invention is not limited to those precise embodiments, and that various other changes and modifications may be made by one skilled in the art without departing from the scope of the invention.
Haensch, Wilfried, Pfeiffer, Dirk, Han, Shu-Jen, Jenkins, Keith A.
| Patent | Priority | Assignee | Title |
| Patent | Priority | Assignee | Title |
| 7301191, | Sep 16 2004 | Etamota Corporation | Fabricating carbon nanotube transistor devices |
| 7433225, | Jul 06 2006 | International Business Machines Corporation | Scalable magnetic random access memory device |
| 8189419, | Jul 06 2009 | CAVIUM INTERNATIONAL; MARVELL ASIA PTE, LTD | Apparatus for nonvolatile multi-programmable electronic fuse system |
| 8379856, | Jun 17 2009 | Empire Technology Development LLC; The Regents of the University of California | Hardware based cryptography |
| 8592955, | Apr 29 2011 | GLOBALFOUNDRIES U S INC | Accurate deposition of nano-objects on a surface |
| 8766258, | Dec 12 2012 | International Business Machines Corporation | Authentication using graphene based devices as physical unclonable functions |
| 9190134, | Apr 08 2009 | SanDisk Technologies LLC | Three-dimensional array of re-programmable non-volatile memory elements having vertical bit lines and a single-sided word line architecture |
| 20060064587, | |||
| 20060255414, | |||
| 20070114613, | |||
| 20080012047, | |||
| 20090213649, | |||
| 20100050901, | |||
| 20100149895, | |||
| 20100171099, | |||
| 20100295025, | |||
| 20120068326, | |||
| 20120183135, | |||
| 20130126830, | |||
| 20130210211, | |||
| 20130232587, | |||
| 20140312248, | |||
| 20140364574, | |||
| 20150221884, | |||
| 20150235053, | |||
| 20160016801, | |||
| 20160087232, |
| Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
| Dec 23 2014 | HAENSCH, WILFRIED | International Business Machines Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 035897 | /0477 | |
| Dec 23 2014 | PFEIFFER, DIRK | International Business Machines Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 035897 | /0477 | |
| Dec 28 2014 | JENKINS, KEITH A | International Business Machines Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 035897 | /0477 | |
| Dec 29 2014 | HAN, SHU-JEN | International Business Machines Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 035897 | /0477 | |
| Jun 24 2015 | International Business Machines Corporation | (assignment on the face of the patent) | / |
| Date | Maintenance Fee Events |
| May 31 2021 | REM: Maintenance Fee Reminder Mailed. |
| Nov 15 2021 | EXP: Patent Expired for Failure to Pay Maintenance Fees. |
| Date | Maintenance Schedule |
| Oct 10 2020 | 4 years fee payment window open |
| Apr 10 2021 | 6 months grace period start (w surcharge) |
| Oct 10 2021 | patent expiry (for year 4) |
| Oct 10 2023 | 2 years to revive unintentionally abandoned end. (for year 4) |
| Oct 10 2024 | 8 years fee payment window open |
| Apr 10 2025 | 6 months grace period start (w surcharge) |
| Oct 10 2025 | patent expiry (for year 8) |
| Oct 10 2027 | 2 years to revive unintentionally abandoned end. (for year 8) |
| Oct 10 2028 | 12 years fee payment window open |
| Apr 10 2029 | 6 months grace period start (w surcharge) |
| Oct 10 2029 | patent expiry (for year 12) |
| Oct 10 2031 | 2 years to revive unintentionally abandoned end. (for year 12) |